circRNA

circRNA applies contextual regression to categorize circular RNAs (circRNAs) by their biogenesis mechanisms using sequence and genomic features.

Key Features:

• Contextual regression model: A machine learning model trained to predict circular RNA formation from random genomic loci on the human genome.
• Feature set: Uses potential biogenesis factors as features, including sequence features within flanking regions of back-spliced junction sites.
• Prediction performance: Achieves high prediction accuracy for circRNA formation.
• Feature extraction: Employs feature extraction techniques to identify sequence- and genome-derived signals associated with circRNA biogenesis.
• Classification into subgroups: Categorizes human circRNAs into seven distinct subgroups based on specific sequence features.
• Sequence features identified: Subgroups are defined by RNA editing sites, simple repeat sequences, self-chains, RNA binding protein binding sites, and CpG islands in flanking regions of back-spliced junction sites.
• Mechanistic insights: Supports coexistence of multiple biogenesis mechanisms and reveals putative correlations between circRNA formation and specific RNA binding proteins and flanking CpG islands.

Scientific Applications:

• Mechanism classification: Classifies circRNAs into mechanistic subgroups to inform studies of circRNA biogenesis.
• Genome-wide prediction: Predicts circRNA-forming loci from random genomic loci on the human genome for genome-wide annotation.
• RBP and CpG island association discovery: Identifies putative correlations between circRNA biogenesis and specific RNA binding proteins and flanking CpG islands.
• Feature-driven hypothesis generation: Provides candidate sequence features—such as editing sites, repeats, and self-chains—for experimental validation of biogenesis mechanisms.

Methodology:

The method employs a contextual regression machine learning model trained on random genomic loci using potential biogenesis factors as features, combined with feature extraction to categorize circRNAs into seven subgroups based on sequence features in flanking regions of back-spliced junction sites.